smart-expense-tracker-cn

Pass

Audited by Gen Agent Trust Hub on Jul 26, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [PROMPT_INJECTION]: The record.sh script lacks proper sanitization for user-provided descriptions, creating a surface for indirect injection into the local database. Ingestion points: Raw text input from the user is processed in record.sh. Boundary markers: The skill does not use specific delimiters or instructions to prevent the agent from misinterpreting content embedded within recorded notes. Capability inventory: The skill writes user input to ~/.expense-tracker/records.jsonl, which is subsequently parsed by jq in summary.sh and query.sh. Sanitization: While the script filters some keywords and limits note length, it does not escape double quotes. A user could provide input containing double quotes to inject additional JSON fields or corrupt the line-delimited JSON format.
  • [COMMAND_EXECUTION]: The skill relies on local shell scripts to manage data, calculate summaries, and configure system tasks. Scripts use common utilities like grep, sed, and cat to manipulate files in the user's home directory. setup-cron.sh facilitates the preparation of a local cron job for scheduled reporting, which requires manual user confirmation for final setup.
  • [EXTERNAL_DOWNLOADS]: The skill requires the jq utility, recommending installation via standard, well-known system package managers such as apt-get or brew.
  • [SAFE]: There is a discrepancy between the scripts listed in SKILL.md (e.g., voice-parse.sh, budget-check.sh) and the scripts actually provided in the package (e.g., daily-push.sh, delete-record.sh). This appears to be a documentation inconsistency rather than a security risk.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 26, 2026, 07:19 PM